A method and system for extracting annual sea ice images based on multiple optical sensors

Abstract

The application provides a kind of annual sea ice image extraction method and system based on multiple optical sensors, it is related to sea ice extraction technical field, based on pixel blue wave band reflectivity, the segmentation threshold between thick ice and thin ice is determined, thick ice image is obtained, then thin ice area is divided into first suspended sediment concentration area image, second suspended sediment concentration area image and third suspended sediment concentration area image according to partition index, sea ice and seawater in similar suspended sediment sea area are divided in the same area, to carry out sea ice extraction respectively, in the face of suspended sediment concentration, still can have higher sea ice segmentation accuracy in estuary nearshore sea area that there is greater change;Based on the maximum between-class variance method, find the segmentation threshold of sea ice and seawater in each level suspended sediment concentration area image, successfully separate thin ice and seawater, combine the thick ice image and thin ice image extracted, efficiently and automatically obtain annual sea ice image.

Description

Technical Field

[0001] This invention relates to the field of sea ice image extraction technology, and in particular to a method and system for extracting images of annual sea ice based on multiple optical sensors. Background Technology

[0002] Annual sea ice is a crucial indicator of mid-latitude climate change, and its distribution and spatiotemporal evolution are an environmental factor that cannot be ignored in marine activities such as nearshore operations, ocean transportation, and fisheries. Currently, commonly used sea ice extent extraction methods can be mainly divided into machine learning methods and spectral index extraction methods. However, machine learning methods often rely on expert prior knowledge and human-computer interaction to select training samples, while spectral index extraction methods are prone to significant errors in nearshore estuarine areas (e.g., the Yellow River estuary) where suspended sediment concentrations vary considerably. These errors are particularly severe in extracting sea ice less than 15 cm thick. Furthermore, these methods have poor portability across different optical sensors. Therefore, there is an urgent need for an efficient and automated method for extracting sea ice images, applicable to common spectral band optical sensors, and with strong portability for annual sea ice extraction. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for extracting images of annual sea ice based on multiple optical sensors. This method achieves efficient and automatic extraction of annual sea ice images and is applicable to optical sensors with four bands: red, green, blue, and near-infrared. It is highly portable.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] On one hand, the present invention provides a method for extracting images of annual sea ice based on multiple optical sensors, comprising the following steps:

[0006] Obtain the sea ice image to be extracted.

[0007] The sea ice image to be extracted is divided into thick ice image and thin ice image based on the thick ice segmentation threshold. The thick ice segmentation threshold is determined based on the blue band reflectance of each pixel in the sea ice image to be extracted.

[0008] Based on the zoning index of each pixel in the thin ice image, images of primary suspended sediment concentration areas, secondary suspended sediment concentration areas, and tertiary suspended sediment concentration areas are extracted from the thin ice image; the zoning index of a pixel is the ratio of the red band reflectance of the pixel to the green band reflectance of the pixel.

[0009] Based on the band reflectance set, images of primary thin ice, secondary thin ice, and tertiary thin ice were extracted from images of primary suspended sediment concentration regions, secondary suspended sediment concentration regions, and tertiary suspended sediment concentration regions, respectively.

[0010] Based on images of thick ice, first-level thin ice, second-level thin ice, and third-level thin ice, images of annual sea ice are obtained.

[0011] Optionally, based on the partition index of each pixel in the thin ice image, images of primary suspended sediment concentration regions, secondary suspended sediment concentration regions, and tertiary suspended sediment concentration regions are extracted from the thin ice image, specifically including:

[0012] Calculate the partition index of each pixel in the thin ice image.

[0013] Regions in the thin ice image where the pixel partitioning index is higher than the first concentration segmentation threshold are identified as first-level suspended sediment concentration regions.

[0014] A secondary thin ice image is obtained by removing the primary suspended sediment concentration area from the thin ice image.

[0015] The second concentration segmentation threshold is determined based on the Otsu's method. The second concentration segmentation threshold is the segmentation threshold for the secondary suspended sediment concentration region and the tertiary suspended sediment concentration region determined by the partition index of each pixel in the secondary thin ice image.

[0016] Based on the second concentration segmentation threshold, images of secondary suspended sediment concentration regions and tertiary suspended sediment concentration regions are extracted from the secondary thin ice images.

[0017] Optionally, based on the band reflectance set, images of primary, secondary, and tertiary thin ice are extracted from the images of primary suspended sediment concentration regions, secondary suspended sediment concentration regions, and tertiary suspended sediment concentration regions, respectively, specifically including:

[0018] Based on the Otsu's method, and according to the band reflectance set, the first-level ice-water separation threshold, the second-level ice-water separation threshold, and the third-level ice-water separation threshold are determined in the images of the first-level suspended sediment concentration region, the second-level suspended sediment concentration region, and the third-level suspended sediment concentration region, respectively.

[0019] Based on the first-level ice-water segmentation threshold, the first-level thin ice image is extracted from the first-level suspended sediment concentration area image.

[0020] Based on the secondary ice-water segmentation threshold, the secondary thin ice image is extracted from the secondary suspended sediment concentration area image.

[0021] Based on the three-level ice-water segmentation threshold, the three-level thin ice image is extracted from the three-level suspended sediment concentration area image.

[0022] Optionally, the band reflectance set includes blue band reflectance and near-infrared band reflectance; based on the maximum inter-class variance method, according to the band reflectance set, the first-level ice-water separation threshold, the second-level ice-water separation threshold, and the third-level ice-water separation threshold are determined in the images of the first-level suspended sediment concentration region, the second-level suspended sediment concentration region, and the third-level suspended sediment concentration region, respectively, specifically including:

[0023] Based on the Otsu's method, the first-level ice-water segmentation threshold is determined according to the blue band reflectance of each pixel in the first-level suspended sediment concentration area image.

[0024] Based on the Otsu's method, the secondary ice-water segmentation threshold is determined according to the near-infrared reflectance of each pixel in the secondary suspended sediment concentration area image.

[0025] Based on the Otsu's method, the three-level ice-water segmentation threshold is determined according to the near-infrared reflectance of each pixel in the three-level suspended sediment concentration area image.

[0026] Optionally, an image of the sea ice to be extracted is acquired, specifically including:

[0027] Acquire nearshore spectral images that include the sea ice images to be extracted.

[0028] Based on the land-sea segmentation threshold, the sea ice image to be extracted is obtained from the nearshore spectral image. The land-sea segmentation threshold is determined based on the maximum inter-class variance method and the normalized water index of each pixel in the nearshore spectral image.

[0029] Optionally, the normalized water index of a pixel is calculated using the following formula:

[0030] NDWI=(RsG-RsNIR) / (RsG+RsNIR).

[0031] Wherein, NDWI is the normalized water index of the pixel, RsG is the green band reflectance of the pixel, and RsNIR is the near-infrared band reflectance of the pixel.

[0032] Optionally, the method further includes:

[0033] Based on the sea ice image to be extracted, the dissimilarity texture features of the sea ice image to be extracted are calculated.

[0034] Based on the dissimilar texture features of the sea ice image to be extracted, the sea ice texture image is obtained.

[0035] Based on the annual sea ice image and the sea ice texture image, an annual sea ice image with sea ice texture is obtained.

[0036] On the other hand, corresponding to the aforementioned method for extracting annual sea ice images based on multiple optical sensors, the present invention also provides a system for extracting annual sea ice images based on multiple optical sensors. When the system is run by a computer, it executes the method for extracting annual sea ice images based on multiple optical sensors as described above.

[0037] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0038] This invention provides a method and system for extracting one-year-old sea ice images based on multiple optical sensors. The method includes: determining the segmentation threshold between thick ice and thin ice based on the blue band reflectance of each pixel in the sea ice image to be extracted, thereby obtaining a thick ice image; then dividing the thin ice region into first-level suspended sediment concentration region images, second-level suspended sediment concentration region images, and third-level suspended sediment concentration region images according to a zoning index; using a threshold segmentation algorithm, such as the maximum inter-class variance method, to find the segmentation threshold between sea ice and seawater in each level of suspended sediment concentration region image, successfully separating the thin ice and seawater; and merging the extracted thick ice image and thin ice image to efficiently and automatically obtain an one-year-old sea ice image. Compared to existing machine learning methods, this invention eliminates the need for prior expert knowledge and human-computer interaction in selecting training samples. It efficiently extracts annual sea ice images simply through automatic extraction of the sea ice images to be extracted. Compared to existing exponential extraction methods, this invention separates sea ice and seawater in similar suspended sediment areas into the same region for separate sea ice extraction. Even in nearshore estuarine areas where suspended sediment concentrations vary significantly, it maintains high sea ice segmentation accuracy. Furthermore, the solution only requires four commonly used bands: red, green, blue, and near-infrared, without the need for additional auxiliary data. It is applicable to all optical sensors containing these four bands, making it highly portable. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a method for extracting images of annual sea ice based on multiple optical sensors, provided in Embodiment 1 of the present invention.

[0041] Figure 2 This is a diagram illustrating the effect of land-sea separation in the method provided in Embodiment 1 of the present invention.

[0042] Figure 3 A comparison diagram of the reflectance of thick ice and thin ice in the method provided in Embodiment 1 of the present invention;

[0043] Figure 4 This is a schematic diagram of the SPM partitioning threshold in the method provided in Embodiment 1 of the present invention;

[0044] Figure 5 This is a flowchart of step A3 in the method provided in Embodiment 1 of the present invention;

[0045] Figure 6 This is a diagram illustrating the division of regions with high suspended sediment concentration, medium suspended sediment concentration, and low suspended sediment concentration in the method provided in Embodiment 1 of the present invention.

[0046] Figure 7 This is a schematic diagram illustrating the method for selecting the optimal image bands for each partition in Embodiment 1 of the present invention.

[0047] Figure 8 This is a schematic diagram illustrating the sea ice extraction results of the method provided in Embodiment 1 of the present invention on the MODIS image of the Bohai Sea region on February 19, 2010, the Landsat 8 image of the Yellow River Estuary on January 23, 2019, and the Sentinel-2 image of the Yellow River Estuary on January 12, 2018.

[0048] Figure 9 This is a schematic diagram of the structure of a one-year-old sea ice image extraction system based on multiple optical sensors provided in Embodiment 2 of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] The purpose of this invention is to provide a method and system for extracting images of annual sea ice based on multiple optical sensors. This method achieves efficient and automatic extraction of annual sea ice images and is applicable to optical sensors with four bands: red, green, blue, and near-infrared. It is highly portable.

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Example 1:

[0053] This embodiment provides a method for extracting images of one-year-old sea ice based on multiple optical sensors, including the following steps:

[0054] S1. Obtain the sea ice image to be extracted. In practical applications, the acquired spectral images often include land images of the nearshore coast. Therefore, before extracting sea ice, it is necessary to separate the land and sea to eliminate the influence of land features on the accuracy of sea ice extraction. In this embodiment, the Normalized Difference Water Index (NDWI) is used to generate a land mask from the nearshore spectral image to achieve land-sea separation. Step S1 specifically includes:

[0055] S11. Obtain a nearshore spectral image that includes the sea ice image to be extracted. It is understood that the nearshore spectral image includes both the ocean image of the sea ice to be extracted and the land image of the nearshore coastline.

[0056] S12. Based on the land-sea segmentation threshold, the sea ice image to be extracted is obtained from the nearshore spectral image. The land-sea segmentation threshold is determined based on the normalized water index of each pixel in the nearshore spectral image, using the maximum inter-class variance method. In this embodiment, the normalized water index of the pixel is calculated using the following formula:

[0057] NDWI=(RsG-RsNIR) / (RsG+RsNIR).

[0058] Wherein, NDWI is the normalized water index of the pixel, RsG is the green band reflectance of the pixel, and RsNIR is the near-infrared band reflectance of the pixel.

[0059] After performing land-sea separation on the nearshore spectral image in step 1, the sea ice image to be extracted is obtained. The land-sea separation result is as follows: Figure 2 As shown, the left side is the original nearshore spectral image, and the right side is the obtained sea ice image to be extracted, with the land portion masked off. Next, the sea ice in the image obtained in step 1 is extracted separately according to industry standards, categorized into thick ice (over 10 cm) and thin ice (under 10 cm). Annual sea ice thickness is generally less than 30 cm. The thick ice mentioned in this method refers to sea ice thicker than 10 cm, mainly including: grey ice, grey-white ice, and white ice; thin ice refers to sea ice thinner than 10 cm, mainly including: new ice, ice rind, and Nile ice.

[0060] S2. Divide the sea ice image to be extracted into thick ice and thin ice images according to the thick ice segmentation threshold; the thick ice segmentation threshold is determined based on the blue band reflectance of each pixel in the sea ice image to be extracted. Figure 3 It is evident that the blue band reflectance of thick ice in the sea ice image to be extracted is significantly higher than that of thin ice. Therefore, the segmentation threshold between thick and thin ice can be directly set to 0.15. Regions with blue band reflectance RsB > 0.15 are classified as thick ice images, and regions with blue band reflectance RsB < 0.15 are classified as thin ice images.

[0061] Thin ice with a thickness of less than 10 cm has a very similar spectral shape to its adjacent seawater. Furthermore, as the suspended sediment concentration (SPM) in the seawater changes, the spectral reflectance peaks of both sea ice and seawater exhibit a "redshift," resulting in drastic changes in both spectral shape and reflectance. Therefore, confusion can easily arise when extracting thin ice and seawater. In this embodiment, a partition index (PI) is proposed based on the spectral characteristics of seawater and sea ice at different suspended sediment concentrations. Adaptive partitioning is achieved through the Otsu's method (maximum inter-class variance) and the seawater reflectance saturation phenomenon. This aims to group seawater with similar suspended sediment concentrations and sea ice located within the same region for separate extraction.

[0062] The phenomenon of seawater reflectance saturation occurs when, as the amount of suspended sediment (SPM) in seawater increases, the blue band reaches saturation first, followed by the green and red bands. Once the blue band reflectance reaches saturation, it no longer increases. At this point, the reflectance values ​​of the red and green bands are approximately equal (i.e., PI = 1), and the concentration of suspended sediment in the seawater is very high, indicating a high concentration of suspended sediment. Figure 4 As shown, the ocean area is divided into three levels: High Suspended Sediment Area (HSSA), Medium Suspended Sediment Area (MSSA), and Low Suspended Sediment Area (LSSA). The images corresponding to these three levels of suspended sediment concentration are further divided into Level 1, Level 2, and Level 3 suspended sediment concentration images.

[0063] S3. Based on the zoning index of each pixel in the thin ice image, extract the primary suspended sediment concentration region image, the secondary suspended sediment concentration region image, and the tertiary suspended sediment concentration region image from the thin ice image; the zoning index of a pixel is the ratio of the red band reflectance to the green band reflectance of the pixel. For example... Figure 5 The flowchart shown includes step S3, which specifically includes:

[0064] S31. Calculate the partition index of each pixel in the thin ice image. The partition index PI is calculated as: red band reflectance RsR / green band reflectance RsG.

[0065] S32. The region in the thin ice image where the pixel partitioning index is higher than the first concentration segmentation threshold is identified as the first-level suspended sediment concentration region image. Figure 4 The first concentration segmentation threshold is 1. Areas with a zoning index (PI) higher than this threshold are considered high suspended sediment concentration areas, and the corresponding images are images of the first-level suspended sediment concentration areas.

[0066] S33. Remove the primary suspended sediment concentration region image from the thin ice image to obtain a secondary thin ice image. Generate a mask for the extracted primary suspended sediment concentration region image and remove it from the thin ice image. Divide the remaining secondary thin ice image into regions with medium suspended sediment concentration and low suspended sediment concentration.

[0067] S34. Based on the Otsu's method, determine the second concentration segmentation threshold; the second concentration segmentation threshold is the segmentation threshold for the secondary suspended sediment concentration region and the tertiary suspended sediment concentration region determined according to the partition index of each pixel in the secondary thin ice image. Figure 4 In this context, the corresponding second concentration segmentation threshold is a1, which is determined based on the Otsu's method of maximum interclass variance. This method can effectively distinguish between regions with medium suspended sediment concentration and regions with low suspended sediment concentration.

[0068] S35. Based on the second concentration segmentation threshold, extract the secondary suspended sediment concentration region image and the tertiary suspended sediment concentration region image from the secondary thin ice image. Based on the second concentration segmentation threshold a1 determined in step S34, the secondary thin ice image can be divided into a medium suspended sediment concentration region and a low suspended sediment concentration region, correspondingly obtaining the secondary suspended sediment concentration region image and the tertiary suspended sediment concentration region image. The segmentation results are as follows: Figure 6 As shown.

[0069] In the primary, secondary, and tertiary suspended sediment concentration area images obtained in step 3, the reflectivity of sea ice is generally higher than that of seawater, which can be determined based on... Figure 7 By selecting the best image bands and comparing the four spectral bands in regions with high suspended sediment concentration, medium suspended sediment concentration, and low suspended sediment concentration, it was found that the blue band had the best segmentation effect in the high suspended sediment concentration region, while the near-infrared band had the best segmentation effect in the medium and low high suspended sediment concentration regions.

[0070] S4. Extract the primary, secondary, and tertiary thin ice images. Specifically, in this embodiment, based on the band reflectance set, the primary, secondary, and tertiary thin ice images are extracted from the primary suspended sediment concentration region images, the secondary suspended sediment concentration region images, and the tertiary suspended sediment concentration region images, respectively. Step S4 specifically includes:

[0071] S41. Based on the Otsu's method, according to the band reflectance set, the first-level ice-water separation threshold, the second-level ice-water separation threshold, and the third-level ice-water separation threshold are determined in the images of the first-level suspended sediment concentration region, the second-level suspended sediment concentration region, and the third-level suspended sediment concentration region, respectively. Specifically, in this embodiment, the band reflectance set includes blue band reflectance and near-infrared band reflectance; step S41 specifically includes:

[0072] S411. Based on the Otsu's method, the first-level ice-water segmentation threshold b1 is determined according to the blue band reflectance of each pixel in the first-level suspended sediment concentration area image.

[0073] S412. Based on the Otsu's method, the secondary ice-water segmentation threshold b2 is determined according to the near-infrared reflectance of each pixel in the secondary suspended sediment concentration area image.

[0074] S413. Based on the Otsu's method, the third-level ice-water segmentation threshold b3 is determined according to the near-infrared reflectance of each pixel in the image of the third-level suspended sediment concentration area.

[0075] S42. Based on the first-level ice-water segmentation threshold b1, extract the first-level thin ice image from the first-level suspended sediment concentration area image.

[0076] S43. Based on the secondary ice-water segmentation threshold b2, extract the secondary thin ice image from the secondary suspended sediment concentration area image.

[0077] S44. Based on the three-level ice-water segmentation threshold b3, the three-level thin ice image is extracted from the three-level suspended sediment concentration area image.

[0078] S5. Based on the thick ice image, the first-level thin ice image, the second-level thin ice image, and the third-level thin ice image, an image of one-year-old sea ice is obtained. In this embodiment, the thick ice image obtained in step 2, the first-level thin ice image, the second-level thin ice image, and the third-level thin ice image extracted in step 4 are logically combined to obtain the image of one-year-old sea ice.

[0079] Because seawater containing suspended sediment shares spectral similarity with some sea ice, the spectral index method cannot completely eliminate the influence of suspended sediment. Furthermore, the extraction results obtained through threshold segmentation contain noise and patches. Therefore, in this embodiment, after obtaining an image of one-year-old sea ice, the method further includes:

[0080] S6. Based on the sea ice image to be extracted, the dissimilarity texture features of the sea ice image to be extracted are calculated. In this embodiment, an image texture feature method based on gray-level co-occurrence matrix (GLCM) is used, which has been widely used in sea ice extraction and can effectively solve the problems of spectral confusion, noise and patchiness that occur in the sea ice detection process; therefore, the accuracy of sea ice extraction can be improved by combining the results of spectral feature and texture feature extraction.

[0081] The specific steps are as follows: 1) GLCM has features such as anisotropy, contrast, entropy, and second moment. This method selects anisotropy features after analyzing the GLCM features of sea ice, with a window size of 10 and an xy shift distance of 1. 2) Calculate the anisotropy texture features of the GLCM method based on the blue band of the marine image in step 1; apply the formula RsB to the anisotropy texture features. diss >60 Extraction of sea ice; RsB diss The blue band reflectance of pixels in an image with heterogeneous texture features.

[0082] S7. Based on the dissimilarity texture features of the sea ice image to be extracted, the sea ice texture image is extracted.

[0083] S8. Based on the annual sea ice image and the sea ice texture image, obtain an annual sea ice image with sea ice texture. Perform a logical AND operation on the annual sea ice image obtained in step S5 and the sea ice texture image extracted in step S7 to obtain the final annual sea ice image with sea ice texture.

[0084] Figure 8 The three columns of images from left to right show the sea ice extraction results of the Bohai Sea region MODIS (500m) image on February 19, 2010, the Yellow River Estuary Landsat 8 (30m) image on January 23, 2019, and the Yellow River Estuary Sentinel-2 (10m) image on January 12, 2018, using the multi-optical sensor-based annual sea ice image extraction method provided in this embodiment. The top image of each column is the original nearshore spectral image, and the bottom image of the corresponding column is the corresponding annual sea ice image extraction result.

[0085] This embodiment provides a method for extracting annual sea ice images based on multiple optical sensors. Compared with existing techniques that use machine learning algorithms, this embodiment's solution does not require prior expert knowledge or human-computer interaction to select training samples. It can efficiently extract annual sea ice images simply by automatically extracting the sea ice images to be extracted. Compared with existing techniques that use spectral index extraction, this embodiment's solution separates sea ice and seawater in similar suspended sediment areas into the same region for separate sea ice extraction. Even in nearshore estuarine waters where suspended sediment concentration varies significantly, it can still achieve high sea ice segmentation accuracy. Furthermore, this embodiment's solution only requires the use of four commonly used bands: red, green, blue, and near-infrared. No other auxiliary data is needed, making it applicable to all other optical sensors containing these four common bands, thus exhibiting strong portability.

[0086] Example 2:

[0087] Furthermore, the method of Embodiment 1 of the present invention can also be used by means of Figure 9 The architecture of the one-year-old sea ice image extraction system based on multiple optical sensors is shown below. Figure 9 As shown, the annual sea ice image extraction system may include a sea ice image acquisition module M1, a thick ice image extraction module M2, a suspended sediment zoning module M3, a thin ice image extraction module M4, and a sea ice image synthesis module M5; some modules may also have sub-units for implementing their functions; for example, the suspended sediment zoning module M3 may also include a zoning index calculation unit, a primary suspended sediment concentration region image extraction unit, a secondary concentration segmentation threshold determination unit, a secondary suspended sediment concentration region image extraction unit, and a tertiary suspended sediment concentration region image extraction unit. Of course, Figure 9 The architecture shown is merely exemplary; it can be omitted as needed when implementing different functionalities. Figure 9 One or at least two components of the system shown.

[0088] Specific examples are used in this article, but the above description is only to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. Those skilled in the art should understand that the various modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, and thus, they can be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.

[0089] Furthermore, those skilled in the art will recognize that, based on the principles of this invention, there will be variations in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A method for extracting images of one-year-old sea ice based on multiple optical sensors, characterized in that, include: Obtain the sea ice image to be extracted; The sea ice image to be extracted is divided according to the thick ice segmentation threshold to obtain thick ice image and thin ice image; The thick ice segmentation threshold is a segmentation threshold for thick ice and thin ice determined based on the blue band reflectance of each pixel in the sea ice image to be extracted. Based on the partition index of each pixel in the thin ice image, images of primary suspended sediment concentration regions, secondary suspended sediment concentration regions, and tertiary suspended sediment concentration regions are extracted from the thin ice image; the partition index of a pixel is the ratio of the red band reflectance of the pixel to the green band reflectance of the pixel. Based on the band reflectance set, the first-level thin ice image, the second-level thin ice image, and the third-level thin ice image are extracted from the images of the first-level suspended sediment concentration area, the second-level suspended sediment concentration area, and the third-level suspended sediment concentration area, respectively. Based on the thick ice image, the first-level thin ice image, the second-level thin ice image, and the third-level thin ice image, an image of one-year-old sea ice is obtained; Based on the partition index of each pixel in the thin ice image, images of primary suspended sediment concentration regions, secondary suspended sediment concentration regions, and tertiary suspended sediment concentration regions are extracted from the thin ice image, specifically including: Calculate the partition index of each pixel in the thin ice image; The regions in the thin ice image whose pixel partitioning index is higher than the first concentration segmentation threshold are identified as the first-level suspended sediment concentration region images; The primary suspended sediment concentration region is removed from the thin ice image to obtain a secondary thin ice image; The second concentration segmentation threshold is determined based on the maximum inter-class variance method; the second concentration segmentation threshold is the segmentation threshold for the secondary suspended sediment concentration region and the tertiary suspended sediment concentration region determined according to the partition index of each pixel in the secondary thin ice image. Based on the second concentration segmentation threshold, secondary suspended sediment concentration region images and tertiary suspended sediment concentration region images are extracted from the secondary thin ice image.

2. The method for extracting images of annual sea ice based on multiple optical sensors according to claim 1, characterized in that, Based on the band reflectance set, primary thin ice images, secondary thin ice images, and tertiary thin ice images are extracted from the images of the primary suspended sediment concentration region, the secondary suspended sediment concentration region, and the tertiary suspended sediment concentration region, respectively. Specifically, this includes: Based on the Otsu's method, according to the band reflectance set, the first-level ice-water separation threshold, the second-level ice-water separation threshold, and the third-level ice-water separation threshold are determined in the images of the first-level suspended sediment concentration region, the second-level suspended sediment concentration region, and the third-level suspended sediment concentration region, respectively. Based on the first-level ice-water segmentation threshold, a first-level thin ice image is extracted from the first-level suspended sediment concentration region image; Based on the secondary ice-water segmentation threshold, a secondary thin ice image is extracted from the secondary suspended sediment concentration region image; Based on the three-level ice-water segmentation threshold, a three-level thin ice image is extracted from the three-level suspended sediment concentration region image.

3. The method for extracting images of annual sea ice based on multiple optical sensors according to claim 2, characterized in that, The band reflectance set includes blue band reflectance and near-infrared band reflectance. Based on the maximum inter-class variance method, according to the band reflectance set, the first-level ice-water separation threshold, the second-level ice-water separation threshold, and the third-level ice-water separation threshold are determined in the images of the first-level suspended sediment concentration region, the second-level suspended sediment concentration region, and the third-level suspended sediment concentration region, respectively. Specifically, these include: Based on the maximum inter-class variance method, the first-level ice-water segmentation threshold is determined according to the blue band reflectance of each pixel in the first-level suspended sediment concentration area image. Based on the maximum inter-class variance method, the secondary ice-water segmentation threshold is determined according to the near-infrared reflectance of each pixel in the secondary suspended sediment concentration area image. Based on the Otsu's method, the three-level ice-water segmentation threshold is determined according to the near-infrared reflectance of each pixel in the three-level suspended sediment concentration area image.

4. The method for extracting images of annual sea ice based on multiple optical sensors according to claim 1, characterized in that, Obtaining images of the sea ice to be extracted specifically includes: Acquire nearshore spectral images including the sea ice image to be extracted; Based on the land-sea segmentation threshold, the sea ice image to be extracted is obtained from the nearshore spectral image; the land-sea segmentation threshold is a segmentation threshold for the sea surface and land determined based on the normalized water index of each pixel in the nearshore spectral image using the maximum inter-class variance method.

5. The method for extracting images of annual sea ice based on multiple optical sensors according to claim 4, characterized in that, The normalized water index of the pixel is calculated using the following formula: NDWI =（ RsG - RsNIR ） / （ RsG + RsNIR ）； in, NDWI The normalized water index for pixels. RsG The green band reflectance of the pixel. RsNIR denoted as the near-infrared reflectance of a pixel.

6. The method for extracting images of one-year-old sea ice based on multiple optical sensors according to claim 5, characterized in that, The method further includes: Based on the sea ice image to be extracted, the dissimilarity texture features of the sea ice image to be extracted are calculated; Based on the dissimilar texture features of the sea ice image to be extracted, a sea ice texture image is obtained; Based on the annual sea ice image and the sea ice texture image, an annual sea ice image with sea ice texture is obtained.

7. A system for extracting images of one-year-old sea ice based on multiple optical sensors, characterized in that, When the multi-optical sensor-based annual sea ice image extraction system is run by a computer, it executes the multi-optical sensor-based annual sea ice image extraction method as described in any one of claims 1-6.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, performs a method for extracting images of annual sea ice based on multiple optical sensors as described in any one of claims 1-6.

Images